TU München - Fakultät für Informatik
Software- and Systems Engineering Research Group

  Home  |  Forschung/Kompetenz  |  Lehre  |  Personen  |  Publikationen  |  Abschlussarbeiten  |  Sonstiges  |  Kontakt  | 

Agenda

Es sprechen Studenten über ihre abgeschlossenen Diplomarbeiten und Systementwicklungsprojekte.

Am Dienstag, 17.09.19, ab 15:30 Uhr, im Raum "Alfred Tarski" (01.11.011B):

Effort Estimation for Code Reviews

Code reviews have been established as an integral part of the software development process. At the same time, effort estimation in software development has yielded promising results in many areas like maintenance or bug-fix time prediction. The question arises, how precise the duration of code reviews can be predicted and what cost factors these predictions depend on. This thesis analyzes work recordings of a commercial software over one and a half years. For completed tasks, review times and potential cost factors were determined. The duration and share of review times was then quantified. It was examined how the cost factors are correlated with the review times and how well review times could be predicted. This was done by training common regression ML models based on the previously computed cost factors. The analysis confirmed that indeed code reviews can be implemented as a lightweight process. Besides, 11% of all development are taken up by code reviews and together with the corresponding rework, this process adds up to 35%. Cost factors quantifying the amount of changes showed the highest correlation with the review time. ML-models did not perform very well on a relative level, but were able to predict review times satisfactorily on an absolute level. The best performing model proved to be a support vector machine with a median absolute error of only 9 minutes.

Hermes: a Protocol for Trading IoT Data Over Distributed Ledgers in Real Time

Internet-of-Things (IoT) sensors have become cheaper and more ubiquitous than ever and this trend seems to be only accelerating for the foreseeable future. From smart thermometers to home assistants and smartphones, massive amounts of data are being streamed in real time by sensors that are continuously connected to the web. However, these data tend to be amassed in private silos that are difficult to access. Private companies and public institutions could build valuable services on top of these data but so far few solutions have been proposed on how to bridge this gap. In this master thesis, we introduce Hermes, a platform for trading sensor data using distributed ledgers as intermediaries to add transparency and safeguards against malicious behavior.